From what I could find in the research, the radius of the fillet should be 6 - 10 % of the chord. There is a small advantage in making the fillet a more complex compound curve in the leading edge area, but the researchers concluded the effort vs benefit was not worthwhile. Most of the benefit occurs in the first third. A fine fillet edge is important, hence the efforts to reduce the gasket width, and finishing the edge to as fine as the material and longevity will allow.
These measures are to increase efficiency and reduce drag by taming pertubed flow patterns in order to make a small surface area fin with improved lifting behaviour, not so much a speed fin. I don't know, but I suspect that at very high speed, the interference pressure waves between fin and board will overwhelm the fillet effect. I am finding it hard to find research on the hydrodynamics of turbine end plates at higher velocities - after all, the turbine engineers are after maximum lifting (power generation) efficiency, not how fast they can get the turbine to spin. I think I will need to follow Yoyo's lead and look at high speed propellor design.
The flip side: the small surface area, the friendly and confidence inspiring handling characteristics of filleted fins could be enough to see them cross the 40 knot mark. I certainly have sailed faster downhill over chop with a filleted fin than with a Delta, simply because I am confident its not going to let go. As Decrepit says, time for some more experiments!